def test_rbm_pseudo_likelihood():
new_rbm = rbm.RBM()
samples = torch.ones(1, 128)
pl = new_rbm.pseudo_likelihood(samples)
assert pl.numpy() < 0
def test_rbm_energy():
new_rbm = rbm.RBM()
samples = torch.ones(1, 128)
energy = new_rbm.energy(samples)
assert energy.numpy() < 0
def test_rbm_visible_sampling():
new_rbm = rbm.RBM()
h = torch.ones(1, 128)
probs, states = new_rbm.visible_sampling(h)
assert probs.size(1) == 128
assert states.size(1) == 128
def test_rbm_hidden_sampling():
new_rbm = rbm.RBM()
v = torch.ones(1, 128)
probs, states = new_rbm.hidden_sampling(v)
assert probs.size(1) == 128
assert states.size(1) == 128
def test_rbm_properties():
new_rbm = rbm.RBM()
assert new_rbm.n_visible == 128
assert new_rbm.n_hidden == 128
assert new_rbm.steps == 1
assert new_rbm.lr == 0.1
assert new_rbm.momentum == 0
assert new_rbm.decay == 0
assert new_rbm.T == 1
assert new_rbm.W.size(0) == 128
assert new_rbm.W.size(1) == 128
assert new_rbm.a.size(1) == 128
assert new_rbm.b.size(1) == 128
def test_rbm_reconstruct():
test = torchvision.datasets.MNIST(
root='./data',
train=False,
download=True,
transform=torchvision.transforms.ToTensor())
test_batches = DataLoader(test,
batch_size=10000,
shuffle=True,
num_workers=1)
new_rbm = rbm.RBM(n_visible=784,
n_hidden=128,
steps=1,
learning_rate=0.1,
momentum=0,
decay=0,
temperature=1)
e, v = new_rbm.reconstruct(test_batches)
assert e >= 0
assert v.size(1) == 784
def test_rbm_fit():
train = torchvision.datasets.MNIST(
root='./data',
train=True,
download=True,
transform=torchvision.transforms.ToTensor())
train_batches = DataLoader(train,
batch_size=128,
shuffle=True,
num_workers=1)
new_rbm = rbm.RBM(n_visible=784,
n_hidden=128,
steps=1,
learning_rate=0.1,
momentum=0,
decay=0,
temperature=1)
e, pl = new_rbm.fit(train_batches, epochs=1)
assert e >= 0
assert pl <= 0